Multidimensional State Data Reduction and Evaluation of College Students’ Mental Health Based on SVM

Abstract

In response to the shortcomings of the traditional methods for evaluating the mental health status of college students in terms of computational complexity and low accuracy, a method for evaluating the mental health status of college students based on data reduction and support vector machines was proposed. A model experiment containing internal and external personality tendency classification, anxiety, and depression dichotomy was designed using logistic regression analysis, information entropy, and SVM algorithm to construct the feature dimensions of the network behavior data, combined with the labeled data of mental state to derive the sample data set for model experiments. In the experimental process, to reflect the difference in the effect of different models, various types of mathematical models were constructed for horizontal comparison; at the same time, to reflect the influence of the parameters of the same type of model, different combinations of parameters were constructed using a grid search algorithm to vertically compare the difference in the effect. The average accuracy of the dichotomous model for anxiety and depression in the sample of 1433 students was 0.80 or higher. The experiments show that the method of predicting students’ psychological status through their online behavioral data is feasible, and the mathematical classification model can be used to grasp students’ psychological status in real time and to warn students with abnormal psychological status, thus helping school counselors to intervene and prevent them promptly.